Molecular dynamics simulation of supercritical fluids

by Branam, Richard D.

Axisymmetric injectors appear in a multitude of applications ranging from rocket engines to biotechnology. While experimentation is limited to larger injectors, much interest has been shown in the micro- and nano-scales as well. Experimentation at these scales can be cost prohibitive if even possible. Often, the operating regime involves supercritical fluids or complex geometries. Molecular dynamics modeling provides a unique way to explore these flow regimes, calculate hard to measure flow parameters accurately, and determine the value of potential improvements before investing in costly experiments or manufacturing. This research effort modeled sub- and supercritical fluid flow in a cylindrical tube being injected into a quiescent chamber. The ability of four wall models to provide an accurate simulation was compared. The simplest model, the diffuse wall, proved useful in getting results quickly but the results for the higher density cases are questionable, especially with respect to velocity profiles and density distributions. The one zone model, three layers of an fcc solid tethered to the lattice sites with a spring, proved very useful for this research primarily because it did not need as many CPU hours to equilibrate. The two zone wall uses springs as a two body potential and has a second stationary zone to hold the wall in place. The most complicated, the three zone wall, employed a reactionary zone, a stochastic zone and a stationary zone using a Lennard-Jones two body potential. Jet simulations were conducted on argon and nitrogen for liquid tube diameters from 20 to 65 Å at both sub and supercritical temperatures (Ar: 130 K and 160 K, N2: 120 K and 130 K). The simulations focused on pressures above the critical pressure (Ar: 6 MPa, N2: 4 MPa). The diffusive wall showed some variation from the analytical velocity profile in the tube while the atomistically modeled walls performed very well. The walls were all able to maintain system temperature to reach the desired simulation conditions. The most dramatic differences between the models were evident in the jet flow into the chamber. The simulation results were strongly influenced by the size of the tube and the wall-fluid interactions. The diffuse wall and the explicitly modeled walls show the ability to compare with macroscaled systems for an issuing jet near the injector region when the iii flow is outside the Knudsen regime. The diffuse wall failed to capture the nanoscale behavior of the flow, in particular at the fluid-wall interface. The explicitly modeled walls performed very well for these locations but determining the appropriate fluid-wall interactions is critical. This interaction is the largest source of error for this research. Comparing the jet results to the available experimental data showed evidence these simulations accurately represent injection flow. Both the diffuse and one zone walls showed good agreement with the density profiles for the larger injection systems. The mass distribution into the chamber compares very well with experimental shadowgraphs. This evidence validates this simulation and suggests further work is possible and appropriate. iv